Linear-Time Data Dissemination in Dynamic Networks

TL;DR

This paper investigates the possibility of achieving linear-time data dissemination in dynamic networks with directed links, improving upon previous bounds and providing new formalism for various network topologies.

Contribution

It introduces a new formalism that simplifies proofs and demonstrates linear-time dissemination under certain structural constraints in dynamic networks.

Findings

Linear-time dissemination is achievable when the number of leaves in rooted trees is bounded.

In rooted chains, data dissemination takes only (n-1) rounds.

For undirected networks, (n-1)/2 rounds suffice in chain graphs.

Abstract

Broadcasting and convergecasting are pivotal services in distributed systems, in particular, in wireless ad-hoc and sensor networks, which are characterized by time- varying communication graphs. We study the question of whether it is possible to disseminate data available locally at some process to all n processes in sparsely connected synchronous dynamic networks with directed links in linear time. Recently, Charron-Bost, F\"ugger and Nowak proved an upper bound of O(n log n) rounds for the case where every communication graph is an arbitrary directed rooted tree. We present a new formalism, which not only facilitates a concise proof of this result, but also allows us to prove that O(n) data dissemination is possible when the number of leaves of the rooted trees are bounded by a constant. In the special case of rooted chains, only (n-1) rounds are needed. Our approach can also be adapted for undirected networks, where only (n-1)/2 rounds in the case of arbitrary chain graphs are needed.